- © 2015 by the Seismological Society of America
Online Material: Relevant accounts for intensity estimates of the 21 October 1880 Portuguese earthquake.
Despite recent important contributions to the completeness of earthquake catalogs, there are still significant gaps in our knowledge about the preinstrumental seismicity of the Iberian Peninsula. This is true even for the second half of the nineteenth century, in spite of the vast increase in the number of scientific institutions and newspapers in these decades. In the last extensive revision of the Iberian earthquake catalog (Rodríguez de la Torre, 1990a,b), the number of known earthquakes occurring between 1851 and 1900 increased from 620 to 2066 as a result of the systematic search conducted by Rodríguez de la Torre mainly using Spanish sources.
Relative to earlier earthquake catalogs, the new data enabled (Martínez Solares and Mezcua, 2002; Mezcua et al., 2004) to present a very much improved catalog of the earthquakes that occurred in the Iberian Peninsula and its vicinity, with a new assessment of epicenters and magnitudes of the major events, through application of the Bakun and Wentworth (B&W) method (Bakun and Wentworth, 1997, 1999; Bakun et al., 2003).
As Martínez Solares and Mezcua (2002) and Mezcua et al. (2004) pointed out, the B&W procedure was only applied to events for which there were enough intensity values. Although registered in this catalog, one major Iberian earthquake that is still waiting for epicenter and magnitude determinations is the 21 October 1880 (6:41 a.m.) earthquake.
The event was unknown until 1990, when it was discovered by Fernando Rodríguez de la Torre (1990a) in Spanish and Portuguese nineteenth century newspapers. With this new information revealed, the 1880 earthquake was included in the catalog of Martínez Solares and Mezcua (2002), which was subsequently transferred into the Spanish Catálogo Sísmico Nacional (Instituto Geográfico Nacional [IGN]) and the SHEEC (SHARE European Earthquake Catalogue, 1000–1899) (Stucchi et al., 2013) with an entry based on Rodríguez de la Torre’s compilation and comprising data from 27 localities. The search was later expanded (Rodríguez de la Torre, 1998) to eventually include macroseismic information from 44 Portuguese and Spanish cities, towns, and villages, although not all are related to the 21 October 1880 event.
Based on the distribution of the intensities, Rodríguez de la Torre (1990a) suggested an Atlantic location “far removed from the coast” and a “high magnitude,” a hypothesis he later reaffirmed (Rodríguez de la Torre, 1998). Martínez Solares and Mezcua (2002) also located the epicenter in the Atlantic Ocean at the approximate coordinates 41.0° N and 10.0° W (i.e., at the latitude of Porto), a location that was also adopted by the SHEEC. In both cases, however, no specific magnitude values were reported, and no calculations were presented to support these hypotheses.
We assume that the number and the spatial distribution of intensity values were not enough to allow the assignment of a reliable epicenter and magnitude to this earthquake through the application of any of the several methods for macroseismic locations now available. Thus, further advances in the knowledge of the macroseismic effects of that event are necessary to determine these source parameters. Finding such additional data and using it to determine the epicenter and the magnitude of the earthquake was the main purpose of our present work.
METHODS AND RESULTS
To compile coeval accounts, a systematic search of the 1880 Portuguese and Spanish national and local press was carried out at a wide range of libraries, archives, and online resources. The search identified more than 100 new accounts of the earthquake, mostly from local Portuguese newspapers. Scientific reports also provided useful data.
The geographic locations of the old and new observations are shown in Figure 1. Macroseismic information is now available from 88 localities (more than tripling the current SHEEC data file); 75 of these are Portuguese, and the vast majority allow the assignment of an intensity value or interval (European Macroseismic Scale 1998 [EMS‐98]). These are supported by a total of 189 independent accounts (available in the electronic supplement to this article). Table 1 summarizes the new set of intensities. In the case of localities already included in the Martínez Solares and Mezcua (2002) dataset, there is a broad agreement between the estimations of intensities of Martínez Solares and Mezcua (2002) and this study.
The new data made it clear that the maximum intensities (VI or even VII in the EMS‐98 scale) were clustered within the country in the vicinity of Serra da Estrela, the highest and longest mountain range in mainland Portugal. Throughout that region, houses and churches collapsed, people were injured, and there was even a death due to a fall (see the electronic supplement). In contrast, in the localities situated along the Atlantic coast of northern and central Portugal, the earthquake caused more fear than damage, with typical intensity values ranging from IV to V (EMS‐98). In southern Portugal and in the Spanish towns from which observations were reported, the earthquake was weakly felt, with intensities rarely exceeding III–IV (EMS‐98). Hence, even without further data handling, the emerging pattern (which is considerably distinct from the one that could be extracted from the previous datasets) made the previous assignments of an Atlantic Ocean epicentral location very unlikely.
To obtain the earthquake location and magnitude using the intensity data, the B&W method was applied. Thereafter, the attenuation law (equation 1) obtained by Martínez Solares and Mezcua (2002) and Mezcua et al. (2004) for the Iberian Peninsula was used to compute the magnitude values Mi for the trial epicenters in accordance with the individual intensity estimates of Table 1 (Ii), being di the distance between the corresponding locality and the trial epicenter. (1)
The trial epicenters were equally spaced (0.1°) and covered an area from 41.5° N to 39.5° N and from 9° W to 7° W. To draw the contour maps, we used the confidence levels tabulated by Bakun and Wentworth (1999) for the number of available intensity values for this earthquake (83).
The map resulting from this procedure is presented in Figure 2, and the corresponding coordinates of the intensity center and value of the moment magnitude (with confidence intervals) are given in Table 2. Our best estimate for the coordinates of the epicentral location, which coincides with the minimum value for the residual root mean square, is 40.42° N and 7.88° W. That places the earthquake source in the inner region of central Portugal, which is almost 200 km away from the previously hypothesized (Martínez Solares and Mezcua, 2002) epicenter in the Atlantic Ocean. The uncertainty associated with this assignment can be assessed (Mezcua et al., 2004) by the radius of the circle of the equivalent area for the different confidence levels: 9, 12, and 26 km for the 50%, 67%, and 95% confidence levels, respectively. By this method, the moment magnitude, obtained from the epicentral value of Mi, was Mw 6.02 (95% CI=5.74–6.21).
DISCUSSION AND CONCLUSIONS
We obtained an initial and reliable estimate of the geographical coordinates for the epicenter of the 21 October 1880 earthquake. In conjunction, we also calculated the magnitude of this seismic event.
A question that inevitably arises from these results is why the researchers who first found and studied the 21 October 1880 event were not able to obtain similar estimates with the intensity data points that they gathered.
The B&W method was developed in 1997, so it was not available to Rodríguez de la Torre for his first analysis (Rodríguez de la Torre, 1990a) nor did he apply it to his subsequent dataset (Rodríguez de la Torre, 1998). However, Martínez Solares and Mezcua (2002) and Mezcua et al. (2004) were successful in the application of the B&W method to 35 earthquakes of the Iberian Peninsula, 40% of them with fewer intensity points than the 22 intensity values that were known for the 21 October 1880 earthquake (Martínez Solares and Mezcua, 2002).
The problem with the earlier epicenter estimates might not be the number of observations but rather their geographic distribution. It is worth noticing that previously there was not a single intensity value from localities situated in the proximity of the epicenter obtained in the present work. Instead, the intensity reports were mostly from the northern and central Portuguese coastal regions and from a few scattered localities in western Spain.
Indeed, the application of the B&W method to that highly biased dataset available prior to the addition of our new data would not have led to reliable estimates of the epicentral location. We applied the B&W method to the earlier dataset and obtained contour lines for the 95% confidence level (95% CI) that extend from 10.4° W to 6.6° W in longitude and from 41.6° N to 39.6° N in latitude. Therefore, the best that the previous investigators could have offered was a guess at the location of the epicenter, which was nevertheless very distant from the earthquake epicenter calculated in the present work.
Despite this, the analysis of the epicenter could have been significantly improved if the B&W method were applied to the second set of observations compiled by Rodríguez de la Torre (1998). In fact, this Spanish investigator presented new findings about the 1880 earthquake in February 1998, at the 4th Portuguese–Spanish Assembly of Geodesy and Geophysics (Rodríguez de la Torre, 1998). Together with a re‐evaluation of his previous intensity estimates, he released macroseismic information for another 17 localities. Unfortunately, his new dataset also included some accounts that (despite being true) are probably misassigned due to confusion of the 21 October 1880 earthquake with the 20 October 1883 event, the latter felt mainly in southern Spain. However, among the valid observations, there were also a handful of newspaper accounts from localities within central Portugal, such as Castelo Novo, Covilhã, and Loriga (see Table 1), which could have given a clue as to the correct location for the earthquake epicenter.
In the present study, we collected data that allowed us to fill some important geographic gaps, particularly within the Portuguese territory. In these more remote locations, news from the local correspondents emerged much later in the newspapers of large urban centers, a circumstance that perhaps was not taken into account in the search by Rodríguez de la Torre (1990a, 1998). In addition, we obtained a number of accounts from local newspapers, most of which are not available in Lisbon’s Biblioteca Nacional de Portugal, where Rodríguez de la Torre completed his inquiries (Rodríguez de la Torre, 1998). Our own circumstances (living in Portugal) allowed easier access to such local sources, and there are now online resources that simply did not exist in 1990 or even in 1998.
In short, we succeeded in our purpose of increasing the initial set of macroseismic data: we advanced the dataset from 27 localities (44 independent accounts) in 1990 to the present total of 88 localities (189 independent accounts). Not only were we able to considerably expand the coverage of the territory, but we were also able to increase the quality of the intensity estimates such that most of them now are based on more than one account.
The estimated coordinates of the epicenter of the 21 October 1880 earthquake, together with the location of the places where the maximum intensities were registered, suggest that the Seia–Lousã fault (Fig. 3), a reverse fault of 90 km length, could be the possible source of this event. This structure is generally considered to have a low seismic activity, as only a few (and relatively small) earthquakes have been assigned to it (Fig. 3).
The importance of the assignment of a magnitude and an epicentral location to this recently discovered historical earthquake goes far beyond a hypothetical new vision of the seismicity of a particular and obscure fault. In fact, the present magnitude value of 6.0 places the 21 October 1880 earthquake as the sixth greatest earthquake of the entire nineteenth century that took place within or near the Iberian Peninsula (Martínez Solares and Mezcua, 2002; Mezcua et al., 2004; Vilanova and Fonseca, 2007; Stucchi et al., 2013). This is a notable achievement for a completely unknown earthquake only 25 years ago.
Without the systematic search conducted by Fernando Rodríguez de la Torre, this earthquake probably would be still forgotten among pages of old newspapers. We are also grateful to him, to José A. Peláez (Universidad de Jáen), to Viviana Castelli (Istituto Nazionale di Geofisica e Vulcanologia), and to Yan B. Radziminovich (Institute of the Earth’s Crust, Irkutsk) for providing scientific papers together with useful comments and encouragement. We acknowledge William H. Bakun (U.S. Geological Survey) for very useful information concerning his method. We also wish to thank Cristina Domingues (Biblioteca do Instituto Geofísico do Infante D. Luís), João Duarte Fonseca (Universidade de Lisboa), José Martínez Solares (Instituto Geográfico Nacional, Spain), and Julio Mezcua (Instituto Geográfico Nacional, Spain) for providing scientific papers. Finally, we are indebted to Arquivo Histórico Municipal de Elvas (Rui Jesuíno), Arquivo Municipal de Huelva (Luísa B. Oliver), Biblioteca de Castilla y León, Valladolid (Amparo S. Rubio), Biblioteca Geral da Universidade de Coimbra (Carlos Fiolhais, José Mateus, and Helena Sousa), Biblioteca Municipal de Beja (Hermes Picamilho), Biblioteca Nacional de Portugal (Área de Reproduções), Biblioteca Pública Municipal do Porto (Paula Bonifácio Marques), Faro de Vigo, and Museu Municipal de Penafiel (Manuel Ribeiro) for helping in the newspaper search.
We are also truly grateful to the anonymous reviewer of our manuscript, whose numerous comments and suggestions greatly helped us to improve the article, and to Claire Niedzwiedz for her work as our language editor, who made significant revisions to the manuscript.